ChatGPT
-
wrote on 24 Jan 2023, 15:35 last edited by
I’m not buying that chatGPT’s bias is entirely pulled from the data. It seems to have some hard coded logic attached to certain people which, for all the efforts of the indoctrinated masses, does not exist across the web to that degree. It’s not in fact impossible to find a nice thing to say about Donald Trump.
-
wrote on 24 Jan 2023, 15:55 last edited by
Does it write pron? That will help move it into the mainstream.
-
I’m not buying that chatGPT’s bias is entirely pulled from the data. It seems to have some hard coded logic attached to certain people which, for all the efforts of the indoctrinated masses, does not exist across the web to that degree. It’s not in fact impossible to find a nice thing to say about Donald Trump.
wrote on 24 Jan 2023, 15:55 last edited byOf course there are manually programmed restrictions. ChatGPT's website tells you upfront there are certain topics that are off limits.
So what if ChatGPT and its maker OpenAI have ideological biases? They are non-governmental entities not obligated to respect your First Amendment rights.
Here's an idea: take GPT-3 and train it with only contemporary American Conservative-leaning data, program in some hard restrictions about topics that contemporary American Conservatives would rather not talk about, and you can make yourself a contemporary American Conservative-leaning chat bot. It's like Conservatopedia as a response to Wikipedia, but with AI!
-
wrote on 24 Jan 2023, 17:23 last edited by
“Is our AIs learning?”
-
wrote on 31 Jan 2023, 02:04 last edited by
-
wrote on 31 Jan 2023, 06:42 last edited by
I wonder if it could solve any of jon’s or klaus’ math riddles.
Never mind…
-
wrote on 31 Jan 2023, 17:40 last edited by
-
wrote on 31 Jan 2023, 17:50 last edited by
@bachophile like any good radiologist, it says that "clinical correlation is required."
-
wrote on 3 Feb 2023, 17:57 last edited by
A 3D printer is used to control a pen to "hand write" homework, using ChatGPT content as the input. Best comment? It's ok, the teacher is going to use OCR and ChatGPT to grade it!
-
-
wrote on 3 Feb 2023, 18:24 last edited by
So for all its sophistication, it's really still failing the Turing test quite badly.
I guess a year from now it won't be, but who knows?
-
wrote on 18 Feb 2023, 03:38 last edited by
Note that ChatGPT has the intelligence to know when to back off and apologize.
-
wrote on 23 Feb 2023, 15:08 last edited by Axtremus
Articles say "ChatGPT-like search" will cost Google multiple $billions in operating cost, "10x" the cost of conventional search. :man-shrugging:
-
wrote on 24 Feb 2023, 13:19 last edited by
-
Of course there are manually programmed restrictions. ChatGPT's website tells you upfront there are certain topics that are off limits.
So what if ChatGPT and its maker OpenAI have ideological biases? They are non-governmental entities not obligated to respect your First Amendment rights.
Here's an idea: take GPT-3 and train it with only contemporary American Conservative-leaning data, program in some hard restrictions about topics that contemporary American Conservatives would rather not talk about, and you can make yourself a contemporary American Conservative-leaning chat bot. It's like Conservatopedia as a response to Wikipedia, but with AI!
wrote on 26 Feb 2023, 23:12 last edited bySo what if ChatGPT and its maker OpenAI have ideological biases? They are non-governmental entities not obligated to respect your First Amendment rights.
.
Here's an idea: take GPT-3 and train it with only contemporary American Conservative-leaning data, program in some hard restrictions about topics that contemporary American Conservatives would rather not talk about, and you can make yourself a contemporary American Conservative-leaning chat bot. It's like Conservatopedia as a response to Wikipedia, but with AI!So, this is happening:
https://www.vice.com/en/article/epzjpn/ai-chatbot-white-supremacist-gab
…
Gab, a white supremacist forum that’s a favorite of mass shooters and organizers of the Capitol riot, and 8kun, the home of QAnon, have announced they’re launching AI engines— and they’re training them on the same content that has in the past led multiple internet companies to cut ties and take the platforms offline.
… -
wrote on 3 Mar 2023, 02:31 last edited by
ChatGPT Is So Bad at Essays That Professors Can Spot It Instantly
“The first indicator that I was dealing with AI was that, despite the syntactic coherence of the essay, it made no sense.”
...
